SyncNet based on Meta's Perception Encoder Audio-Visual (PE-AV)
-
Updated
Jul 28, 2026 - Python
SyncNet based on Meta's Perception Encoder Audio-Visual (PE-AV)
LTX-2.3 talking-head UGC video pipeline with measurement gates: SyncNet lip-sync and WER gates, script lint, best-of-K seed banks, a scoring fleet (UTMOS, NISQA, ECAPA, ArcFace, RAFT, LPIPS) and the 22-experiment autoresearch leaderboard
An experiment testing Wav2Lip/SyncNet lip-sync robustness under noisy audio conditions.
Public source repository for SyncNet — an application layer for crypto-native project creation and experimentation.
To associate your repository with the syncnet topic, visit your repo's landing page and select "manage topics."